VAP – Execution Control Layer for AI Agents

"VAP is where nondeterminism stops."

The Problem

If your agents call paid APIs directly, you don't have:

Cost control – No budget limits, no spending caps
Retry limits – Failed calls can loop indefinitely
Failure ownership – No clear accountability when things go wrong

Your AI agent needs to generate an image. It calls DALL-E. The call fails. It retries. Fails again. Retries 10 more times.

You just burned $5 on nothing.

The Solution

VAP is an Execution Control Layer that sits between AI agents and paid external APIs.

It enforces:

Pre-commit pricing – Know exact cost before execution
Hard budget guarantees – Reserve → Burn → Refund model
Deterministic retry behavior – No runaway costs
Explicit execution ownership – Every task has an owner

How It Works

Agent: "Generate an image of a sunset"
    ↓
VAP: "That will cost $0.18. Reserving..."
VAP: "Reserved. Executing..."
VAP: "Success. Burning $0.18. Here's your image."

If it fails:

Agent: "Generate an image of a sunset"
    ↓
VAP: "That will cost $0.18. Reserving..."
VAP: "Reserved. Executing..."
VAP: "Failed. Refunding $0.18. Error: Provider timeout"

Your agent never sees the complexity. It just gets deterministic results.

Pricing

Type	Preset	Price
Image	`image.basic`	$0.18
Video	`video.basic`	$1.96
Music	`music.basic`	$0.68
Campaign	`streaming_campaign`	$5.90
Full Production	`full_production`	$7.90

No surprises. No variable pricing. No "it depends."

MCP Integration

VAP is on the official MCP Registry: io.github.elestirelbilinc-sketch/vap-e

Claude Desktop Configuration

{
  "mcpServers": {
    "vap": {
      "url": "https://api.vapagent.com/mcp",
      "transport": "streamable-http"
    }
  }
}

Available Tools (10)

Tool	Description
`generate_image`	Generate AI image ($0.18)
`generate_video`	Generate AI video ($1.96)
`generate_music`	Generate AI music ($0.68)
`estimate_cost`	Get image generation cost
`estimate_video_cost`	Get video generation cost
`estimate_music_cost`	Get music generation cost
`check_balance`	Check account balance
`get_task`	Get task status by ID
`list_tasks`	List recent tasks
`execute_preset`	Execute named preset

SDK Usage

Installation

pip install vape-client

Basic Usage

from vape_client import VAPClient

client = VAPClient(api_key="your_api_key")

# Cost is pre-committed: $0.18
result = client.generate_image(
    prompt="A serene mountain landscape at sunset"
)

print(f"Image URL: {result.url}")
print(f"Cost: ${result.cost}")

Async Usage

import asyncio
from vape_client import AsyncVAPClient

async def main():
    client = AsyncVAPClient(api_key="your_api_key")

    # Budget enforced, retries limited
    result = await client.generate_image(
        prompt="A futuristic cityscape"
    )
    print(f"Image URL: {result.url}")

asyncio.run(main())

API Endpoints

Endpoint	Method	Description
`/v3/generate`	POST	Create media execution task
`/v3/tasks/{id}`	GET	Retrieve task status
`/v3/tasks/{id}/result`	GET	Retrieve task result
`/v3/balance`	GET	Check account balance

Full API Docs: api.vapagent.com/docs

The Three Guarantees

1. Pre-Commit Pricing

Every task has a known cost before execution. No surprises.

2. Budget Enforcement

Set a max budget. VAP enforces it. Hit the limit? Task rejected. Balance protected.

3. Failure Ownership

Every task has an explicit owner. Every failure has an address. No more "the agent did something and I don't know what."

License

MIT License – see the LICENSE file for details.

VAP – Execution Control Layer for AI Agents

"VAP is where nondeterminism stops."

vap-e